Leveraging Model-Based Trees as Interpretable Surrogate Models for Model Distillation

MCML Authors

Julia Herbinger

Dr.

* Former Member

→ Group Bernd Bischl
 Statistical Learning and Data Science

Susanne Dandl

Dr.

* Former Member

→ Group Bernd Bischl
 Statistical Learning and Data Science

Fiona Katharina Ewald

→ Group Bernd Bischl
 Statistical Learning and Data Science

Giuseppe Casalicchio

Dr.

→ Group Bernd Bischl
 Statistical Learning and Data Science

Abstract

Surrogate models play a crucial role in retrospectively interpreting complex and powerful black box machine learning models via model distillation. This paper focuses on using model-based trees as surrogate models which partition the feature space into interpretable regions via decision rules. Within each region, interpretable models based on additive main effects are used to approximate the behavior of the black box model, striking for an optimal balance between interpretability and performance. Four model-based tree algorithms, namely SLIM, GUIDE, MOB, and CTree, are compared regarding their ability to generate such surrogate models. We investigate fidelity, interpretability, stability, and the algorithms’ capability to capture interaction effects through appropriate splits. Based on our comprehensive analyses, we finally provide an overview of user-specific recommendations.

inproceedings HDE+23

---

ECAI 2023

3rd International Workshop on Explainable and Interpretable Machine Learning co-located with the 26th European Conference on Artificial Intelligence. Kraków, Poland, Sep 30-Oct 04, 2023.

---

Authors

J. Herbinger • S. Dandl • F. K. Ewald • S. Loibl • G. Casalicchio

---

Links

DOI

---

Research Area

 A1 | Statistical Foundations & Explainability